The RX Vega 56 Owners Thread

[henson0115]

Seems to do this sometimes for me in Overwatch also.

I have tried that as well, it is understandable when the gpu idles for a bit and still hits 90 fps, but when it doesn't the clocks should go up... Other than that, it should always be on the highest state in vr applications.

edit: in the SteamVR frame timing graphs I see the GPU taking up more than 11ms to render frames (for 90fps) and still chilling at 800mhz. When turning the view to more demanding areas the clock actually goes down to 600-500 most of the time... :S

could be an early driver issue, but the problem resolves itself if you turn the setting sup, after all why buy an expensive gpu to keep the settings low, turn em up.

could be an early driver issue, but the problem resolves itself if you turn the settings up, after all why buy an expensive gpu to keep the settings low, turn em up.

 

[lowdog]

So I tried the 64 liquid cooled bios last night and my card wasn't stable at its default clocks so rolled back to the 64 air bios.
After testing im gonna be running my card at 1622mhz @ 1050mV and 1100mhz HBM. Boosts to 1600mhz during Timespy and had a speak power usages of 215w reported by gpuz.
I was running at 1702mhz @ 1200mV with HBM @ 1100mhz. Highest core boost I saw was 1680mhz if i remember right but card would use 309w in Timespy and only gain 200 graphics score points.

Thats a lot of extra power draw for 80mhz!

Yeah my 56 wasn't stable at default/balanced on the 64 AIO Liquid bios either though it would do 720MHz on that bios with 1250mv drawing 340W so I flashed back to Air 64 bios and it's perfectly stable on that both balanced and turbo......I've settled on PL +50% - S6 @ 1542MHz with 1025mv and S7 @ 1637MHz with 1075mv - mem @ 1050MHz with 950mv for an in game steady frequency of 1610MHz with a 220/230W PD.....that's my 24/7 for gaming.

 

[old gamer]

Yeah my 56 wasn't stable at default/balanced on the 64 AIO Liquid bios either though it would do 720MHz on that bios with 1250mv drawing 340W so I flashed back to Air 64 bios and it's perfectly stable on that both balanced and turbo......I've settled on S6 @ 1542MHz with 1025mv and S7 @ 1637MHz with 1075mv - mem @ 1050MHz with 950mv for an in game steady frequency of 1610MHz with a 220/230W PD.....that's my 24/7 for gaming.

what sort of temps did you get on turbo, with the 64 bios and turbo mine hit 85c benching

 

[lowdog]

what sort of temps did you get on turbo, with the 64 bios and turbo mine hit 85c benching

I have a EK full cover block on mine in a custom loop with 2 x 360 rads for 1800X @ 3.9GHZ and the Vega.....the Vega hits anywhere from mid 30C's to mid 40C's depending on intensity of load with 28C ambient room temps....hot spot hits low 50C's max on turbo. With the settings I posted above those temps are lower and hot spot is in the 40C's but doesn't go over 50C max.

Card is inefficient in both balanced and turbo modes....try to dial in your own settings focusing on 1600MHz stable frequency while gaming with a PL + 50% and lowest volts you can achieve...most should do 1600MHz core stable with under 1100mv....just check my settings from a couple of posts back. Leave P6 and P7 frequencies alone and just push the PL to +50% and start lowering the volts for both those states to find the balance between frequencies and voltages for stability. The mem volts don't actually affect the HBM itself but bring down the floor for the volts to the gpu core so if you set 900mv for the mem/HBM you can lower S6 and S7 to 900mv and it will go down that low to the core.

 

[old gamer]

I have a EK full cover block on mine in a custom loop with 2 x 360 rads for 1800X @ 3.9GHZ and the Vega.....the Vega hits anywhere from mid 30C's to mid 40C's depending on intensity of load with 28C ambient room temps....hot spot hits low 50C's max on turbo. With the settings I posted above those temps are lower and hot spot is in the 40C's but doesn't go over 50C max.

nice, getting these cards under water looks the best bet, with undervolting i managed to get the temps mid 70's so not too bad

 

[mobius0]

could be an early driver issue, but the problem resolves itself if you turn the settings up, after all why buy an expensive gpu to keep the settings low, turn em up.

I did, it doesn't solve anything :S clocks stay there and fps drops bellow 90 half the time.

 

[Shiari]

I did, it doesn't solve anything :S clocks stay there and fps drops bellow 90 half the time.

What's the temps and utilisation like?

 

[henson0115]

I did, it doesn't solve anything :S clocks stay there and fps drops bellow 90 half the time.

have you tried a DDU and reinstall? most issues seem to come from simply updating them rather than clean installing them.

 

[mobius0]

What's the temps and utilisation like?

have you tried a DDU and reinstall? most issues seem to come from simply updating them rather than clean installing them.

Usage well bellow 90% most of the time but fluctuating from like 40 to 80, temperatures in the 60s.
Reinstalled a bunch of stuff, used DDU (had a 7970 before that worked properly and even performed better in those occasions).
Clocks can go even lower (500-600) when pointing the view to areas that require more gpu power resulting in very obvious fps drops and choppiness in VR. I don't get it... does its throttling get triggered by something stupid or does it not even go to the highest state?

I am guessing I will get a full refund if I return it within the 14 days right? It is not doing the only thing I bought it for which is VR, I have a 1080 monitor that my 7970 handled just fine before. And even if it is a driver problem I can't wait because I have online league races, although if it was I would expect many more people to complain online, I have only found 1 more case: https://community.amd.com/thread/219535

 

[T.Kay]

Anyone know how to edit the powerplay tables?

 

[henson0115]

Usage well bellow 90% most of the time but fluctuating from like 40 to 80, temperatures in the 60s.
Reinstalled a bunch of stuff, used DDU (had a 7970 before that worked properly and even performed better in those occasions).
Clocks can go even lower (500-600) when pointing the view to areas that require more gpu power resulting in very obvious fps drops and choppiness in VR. I don't get it... does its throttling get triggered by something stupid or does it not even go to the highest state?

I am guessing I will get a full refund if I return it within the 14 days right? It is not doing the only thing I bought it for which is VR, I have a 1080 monitor that my 7970 handled just fine before. And even if it is a driver problem I can't wait because I have online league races, although if it was I would expect many more people to complain online, I have only found 1 more case: https://community.amd.com/thread/219535

if i have missed it then apologies, but what are the rest of your specs? my usuage is 100% in dcs vr, finally got a blooming replacement hdmi cable..
edit:also what power options profile in the control panel are you running out of interest?

 

[Nerusy]

Been tinkering with the overclock and came to a conclusion that firestrike is not a reliable bench. I found perfect spot at +35 power, 1v on both stages and memory at 900mhz resulting in good temps, power draw etc and giving 22,300 graphics score but in games the clocks are stuck at 1200 so actual gaming performance is worse.

I don't get why it wont go above 1200 in games, temps have more than enough room to go, the card can draw more power as evident from firestrike score but it just refuses to go above 1200

Edit: as it turns out setting voltage below 1050mv makes it so firestrike can push the card to max usage with watt usage ~280 whereas in games it the card gives up and doesnt go above 1200mh, usage is spotty and rarely above 180 Watts

Edit2: watch out guys, higher firestrike result may not translate into better game performance on vega.

 

[Spuddy]

is there any info on non-reference cards yet?

 

[mobius0]

if i have missed it then apologies, but what are the rest of your specs? my usuage is 100% in dcs vr, finally got a blooming replacement hdmi cable..
edit:also what power options profile in the control panel are you running out of interest?

Ryzen 1700, AX370 gaming 5, 16Gb 3200 ram, SSD, Win 10. Tried all power options, might give the 2nd bios a try tomorrow. The clocks boost at random points for me, but overall it is so bad that the 7970 looked better. It happens in Dcs, iracing, assetto corsa. Clocks seem to boost properly for steamvr home, not suck at 1300+ but they will go up before missing any frames.

Are your clocks always maxed when in vr applications? Also, if you go crazy with settings and supersampling, does reprojection work properly?

 

[henson0115]

Ryzen 1700, AX370 gaming 5, 16Gb 3200 ram, SSD, Win 10. Tried all power options, might give the 2nd bios a try tomorrow. The clocks boost at random points for me, but overall it is so bad that the 7970 looked better. It happens in Dcs, iracing, assetto corsa. Clocks seem to boost properly for steamvr home, not suck at 1300+ but they will go up before missing any frames.

Are your clocks always maxed when in vr applications? Also, if you go crazy with settings and supersampling, does reprojection work properly?

what psu have you got? its not 100% boosted all the time, but almost especially with an undervolt of 1020. rather than using the profiles in wattman, try cicking reset in wattman and then selecting custom and testing stock speeds but changing your voltage to say 1150 and 1100 rather than stock 1200 and 1150 and see if thats any better for you, the lower you go the more boost you should see.

 

[henson0115]

Been tinkering with the overclock and came to a conclusion that firestrike is not a reliable bench. I found perfect spot at +35 power, 1v on both stages and memory at 900mhz resulting in good temps, power draw etc and giving 22,300 graphics score but in games the clocks are stuck at 1200 so actual gaming performance is worse.

I don't get why it wont go above 1200 in games, temps have more than enough room to go, the card can draw more power as evident from firestrike score but it just refuses to go above 1200

Edit: as it turns out setting voltage below 1050mv makes it so firestrike can push the card to max usage with watt usage ~280 whereas in games it the card gives up and doesnt go above 1200mh, usage is spotty and rarely above 180 Watts

Edit2: watch out guys, higher firestrike result may not translate into better game performance on vega.

try reseting your wattman overclock, and reapplying.

 

[SHOGUNXSHINOBI]

is there any info on non-reference cards yet?

https://videocardz.com/amd

 

[Nerusy]

try reseting your wattman overclock, and reapplying.

Tried that but then something strange started happening, the system would now freeze at stock. This is also frensh windows install. Will need to tinker more but unstable stock scared me yesterday

 

[lowdog]

Tried that but then something strange started happening, the system would now freeze at stock. This is also frensh windows install. Will need to tinker more but unstable stock scared me yesterday

It's the driver and wattman, sometimes the "apply" button just doesn't apply it all correctly and the whole thing becomes unstable ant the system can just freeze opening up an app/gpu-z/3DMark etc. Sometimes the wattman "reset" and 'apply" just don't take and you need to reboot.....pain in the (_!_) but you need to do it and then hit wattman "reset" twice.....just buggy driver/wattman.

 

[mobius0]

what psu have you got? its not 100% boosted all the time, but almost especially with an undervolt of 1020. rather than using the profiles in wattman, try cicking reset in wattman and then selecting custom and testing stock speeds but changing your voltage to say 1150 and 1100 rather than stock 1200 and 1150 and see if thats any better for you, the lower you go the more boost you should see.

Corsair CX750. The card boosts properly in 2D games, although undervolting seems to just drop the clocks a bit as well unless the power limit is raised. Do your clocks go up once you undervolt without touching the power limit? I don't know what might be wrong, there could be a problem when the card outputs on a second display (Vive).

edit: If it was a power limit problem the temperature wouldn't be in the 60s I guess, in 2D games it will boost and go up to 80.